Research

Is Quarto solved? What a solver says about perfect play

By Alejandro Rioja · September 15, 2026 · Raw data (JSON)

Yes. Classic Quarto was solved by computer in 1998 by Luc Goossens. With perfect play from both players it's a draw, and nobody can force a win from the empty board. We wanted to know what happens when 2×2 squares also count as wins, as they do in Quads and in Quarto's advanced variant. So we solved 13,224 random endgames exactly under both rule sets. With seven pieces on the board, draws fell from 42.9% to 5.8% once squares counted.

About us: we make Quads, an independent game built on the same hand-your-opponent-the-piece mechanic. It isn't affiliated with Gigamic or the makers of Quarto. Everything below was computed with our own solver on our own game engine. The method and its limits are spelled out, and the raw numbers are published.

What "solved" means

A game is solved when a computer has worked out the result with perfect play from both sides. For Quarto the answer is a draw. Players often misread this in two ways:

Our question: what do 2×2 squares change?

Classic Quarto wins on 10 shapes: 4 rows, 4 columns and 2 diagonals. Many editions include an advanced variant where any 2×2 square of four matching pieces also wins, and Quads always plays that way. That adds 9 winning shapes, for 19 in total. Intuitively that should make draws rarer. We measured how much.

How we tested it

An important limit: random positions aren't the positions strong players reach. Careful players avoid a lot of what random placement creates. Read these numbers as a comparison of the two rule sets on identical positions, not as the odds of winning your next game. We also didn't re-solve the full game from an empty board with squares.

Finding 1: squares turn draws into decisive games

Here's the result for the player to move with perfect play by both sides, on the same positions under each rule set:

Pieces placedPositionsClassic: winClassic: drawWith squares: winWith squares: draw
424*0%100%45.8%54.2%
5200*18.0%82.0%61.5%37.5%
61,00046.9%53.1%85.3%13.8%
72,00056.3%42.9%91.3%5.8%
82,00075.8%23.4%93.0%3.6%
102,00087.0%10.6%95.7%1.7%
122,00088.9%7.4%96.5%1.1%

* Small samples. Positions with few pieces placed take far longer to solve exactly (with 4 pieces placed, about 12 seconds each), so read those rows as a direction, not a precise rate. The remaining share in each row is a forced loss for the player to move. Results for 9 and 11 pieces placed are in the raw data.

Three things stand out:

Finding 2: safe pieces run out much faster with squares

A piece is safe to hand over if your opponent can't win with it anywhere on the board. We counted safe pieces in 20,000 random positions at each stage:

Pieces placedClassic: pieces safeSquares: pieces safeClassic: no safe pieceSquares: no safe piece
492.4%86.6%0%0%
670.6%56.5%0.5%2.5%
756.9%40.2%3.0%11.0%
843.8%26.7%10.0%27.5%
1024.5%10.4%37.1%66.1%
1216.6%5.3%59.8%85.0%

With eight pieces placed and squares counting, more than a quarter of random positions leave no safe piece at all, and the player who has to hand one over has already lost. Under classic rules the same happens in one position in ten. That's why keeping a safe piece in reserve is the core skill in our strategy guide.

Finding 3: how the Quads AI does against itself

We also had the computer opponents that ship in Quads play each other: 5,000 games per setting, 20,000 in total. The AI makes the same move every time in a given position, so we randomised the opening turns to get varied games. Every game uses 2×2 squares. "First placer" is the player who places the first piece, which means they received the very first hand-off.

LevelRandom opening turnsFirst placer winsOther player winsDraws
Grandmaster vs Grandmaster253.0%19.4%27.6%
Grandmaster vs Grandmaster448.5%29.9%21.6%
Skilled vs Skilled240.8%40.3%18.9%
Skilled vs Skilled442.9%42.9%14.1%

Games averaged about 12.5 placements, so most end with three or four pieces unplayed. Grandmaster's games draw more often than Skilled's, and they lean clearly toward the player who places first. Skilled's games are almost exactly even. We haven't worked out why yet, so treat it as an observation about our current AI, not a fact about perfect play. The AI only searches exactly in the last part of the game and uses rules of thumb before that. Also, with just two random opening turns some openings repeat, so that row is less varied than 5,000 games suggests.

What this means when you play

Frequently asked questions

Is Quarto solved?

Yes. Luc Goossens solved classic Quarto by computer in 1998: with perfect play from both players, the game is a draw. Neither player can force a win from the empty board.

Does the first player have an advantage in Quarto?

Not with perfect play, because the solved result is a draw. In practice, games between imperfect players are usually decided by who is first forced to hand over a winning piece.

Does the 2×2 square variant change the result?

Our study didn't re-solve the full game with squares, but squares change endgames a lot. In 2,000 random positions with seven pieces placed, 42.9% were draws under classic rules and only 5.8% with 2×2 squares counting.